An exonic splicing enhancer offsets the atypical GU-rich 3' splice site of human apolipoprotein A-II exon 3.

Human apolipoprotein A-II (apoA-II) intron 2/exon 3 junction shows a peculiar tract of alternating pyrimidines and purines (GU tract) that makes the acceptor site deviate significantly from the consensus. However, apoA-II exon 3 is constitutively included in mRNA. We have studied this unusual exon definition by creating a construct with the genomic fragment encompassing the whole gene from apoA-II and its regulatory regions. Transient transfections in Hep3B cells have shown that deletion or replacement of the GU repeats at the 3' splice site resulted in a decrease of apoA-II exon 3 inclusion, indicating a possible role of the GU tract in splicing. However, a 3' splice site composed of the GU tract in heterologous context, such as the extra domain A of human fibronectin or cystic fibrosis transmembrane conductance regulator exon 9, resulted in total skipping of the exons. Next, we identified the exonic cis-acting elements that may affect the splicing efficiency of apoA-II exon 3 and found that the region spanning from nucleotide 87 to 113 of human apoA-II exon 3 is essential for its inclusion in the mRNA. Overlapping deletions and point mutations (between nucleotides 91 and 102) precisely defined an exonic splicing enhancer (ESEwt). UV cross-linking assays followed by immunoprecipitation with anti-SR protein monoclonal antibodies showed that ESEwt, but not mutated ESE RNA, was able to bind both alternative splicing factor/splicing factor 2 and SC35. Furthermore, overexpression of both splicing factors enhanced exon 3 inclusion. These results show that this protein-ESE interaction is able to promote the incorporation of exon 3 in mRNA and suggest that they can rescue the splicing despite the noncanonical 3' splice site.

Pre-mRNA splicing is the process by which introns are removed and exons are joined together by a two-step trans-esterification reaction carried out by the spliceosome, a dynamic 60 S ribonucleoprotein particle (1). Formation of the spliceosome at particular splice junctions is triggered by recognition of the 5Ј splice site by the U1 small nuclear ribonucleoprotein and of the 3Ј splice site by U2AF followed by the U2 small nuclear ribonucleoprotein recognition of the branch point (2).
The polypyrimidine tract is one of the important cis-acting sequences present in the 3Ј splice site of introns. The progressive deletion of the polypyrimidine tract abolish lariat formation, spliceosome assembly, and consequently the splicing process (3,4), whereas increasing the length of the pyrimidine run can lead to improved efficiency of splicing in some systems (4).
In vitro studies have demonstrated that the ability of polypyrimidine tracts to favor specific 3Ј splice site selection is not only determined by its length but also by its composition (4,5). This feature coexists with a degree of flexibility in the specific sequence of a given tract.
The human apolipoprotein A-II (apoA-II) 1 gene presents a peculiar arrangement of GT repeats within the polypyrimidine tract region at the intron 2/exon 3 junction that deviates significantly from the consensus (6). This characteristic GT tract is also found within the intron 8/exon 9 junction of the human CFTR gene, but in this context, it is followed by a polymorphic poly(T) tract (7)(8)(9).
This apparent sequence similarity concerning the GU tract is contrasted by the different splicing pattern exhibited by the two genes. In fact, CFTR exon 9 undergoes alternative splicing, and its inclusion is inversely correlated with the length of the GU tract and directly proportional to the length of the poly(T) tract, (7,10) whereas apparently apoA-II exon 3 is constitutively spliced, and its inclusion is dependent on the presence of the GU tract (11). In other words, in the CFTR intron 8/exon 9 context, the stretch of pyrimidines alternated with purines alone is not equivalent to a functional continuous polypyrimidine tract (10), in contrast to what has been observed for the apoA-II gene (11). Altogether these observations prompted us to investigate the mechanisms underlying the constitutive splicing of apoA-II exon 3 and, in particular, to characterize the cis-acting elements and the trans-acting factors involved in apoA-II exon 3 definition.
We show here that the third exon of apoA-II contains an exonic splicing enhancer (ESE), which is essential for the inclusion of exon 3 in mRNA and which possibly balances the presence of the noncanonical 3Ј splice site. The trans-acting factors that target the apoA-II exon 3 ESE have been identified as ASF/SF2 and SC35.

EXPERIMENTAL PROCEDURES
Constructs-To generate the pApo gene system, the full sequence of the gene was obtained by overlapping the PCRs of different fragments from the apoA-II gene (GenBank TM accession number X04898.1) and then cloned XhoI-SacII in pBluescript SK. Each fragment of the complete apoA-II gene was amplified by PCR (94°C for 30 s, 60°C for 30 s, 72°C for 30 s, 30 cycles) using ϳ200 -400 ng of genomic DNA as the template. The primer sequences at the 5Ј and 3Ј extremes of the apoA-II gene were: hapoA-II-1173 XhoI 5Ј, 5Ј-tccgctcgagcccgggaggtggaggttgca-3Ј and hapoA-II poly(A) SacII 3Ј, 5Ј-ttccccgcgggtaggagactctgggtttgga-3Ј. After amplification, the PCR products were purified through a Mi-croSpin S-400HR column (Amersham Biosciences AB, Uppsala, Sweden). At position 1549 (intron 2) and 2017 (intron 3), the SalI and EcoRI sites were introduced by PCR-mediated site-directed mutagenesis, respectively, to facilitate subsequent cloning procedures. Also a target sequence was included in the apoA-II expression system at the end of exon 4 to allow the specific amplification and analysis of transfected apoA-II RNA with specific primers.
The CF/apoA-II hybrids were generated by exchanging the SalI-EcoRI cassettes or through a two-step PCR overlap extension. The three constructs with overlapping 21-bp deletions within the exon 3 were generated by exchanging the SalI-EcoRI cassettes in the pApo gene system.
The introduction of artificial point mutations within the ESE in apoA-II exon 3 were carried out by PCR-mediated site-directed mutagenesis. The amplified fragments were digested by SalI-EcoRI and substituted with the appropriate SalI-EcoRI cassettes created in the previously described pApo gene system.
Before expression, the identity of all constructs was checked by a CEQ2000 sequencer (Beckman Coulter, Fullerton, CA) according to the manufacturer's instructions. The wild type polypyrimidine tract of the EDA exon in the ␣-globin/fibronectin reporter system (12) was replaced by (GT) 16 repeats. Sense and antisense oligos carrying a (GT) 16 were designed such that they annealed at the flanking sequences of the EDA polypyrimidine tract. A first PCR reaction (94°C for 30 s, 60°C for 30 s, 72°C for 30 s, for 30 cycles) was performed by using a sense oligo Ex-1S, 5Ј-atcaaacagaaatgaccattga-3Ј annealing at Ϫ1 EDA exon and the antisense oligo FN-AII rev, 5Ј-atcaatgtctgcacacacacacacacacacacacacacacacactctgttgtg-3Ј carrying the (GT) 16 sequence. A second reaction (same PCR conditions) was carried out by using the sense oligo FN-AII dir, 5Ј-cacaacagagtgtgtgtgtgtgtgtgtgtgtgtgtgtgtgtgcagacattgat-3Ј carrying the (GT) 16 sequence and the antisense oligo Exϩ1AS, 5Ј-ctctttcgggttcacccgca-3Ј annealing at ϩ1 EDA exon. Afterward, a third PCR reaction was carried out (same conditions) by using the primers annealing at the extremes Ex Ϫ 1S and Ex ϩ 1AS and as a template, 1 l each of reactions 1 and 2. The final PCR product was visualized in a 1% agarose gel, purified, cloned, and the plasmid transfected in Hep3B cells.
Transfections-The DNA used for transfections was prepared with JetStar purification kit (Genomed, GmbH, Löhne, Germany) following the manufacturer's instructions. Liposome-mediated transfections of 3 ϫ 10 5 human hepatocarcinoma Hep3B cells were performed using DOTAP liposomal transfection reagent (Alexis Corporation, Lausanne, Switzerland). 3 g of construct DNA were mixed with 5 g of DOTAP reagent for each transfection, and the mixture was incubated at room temperature for 15 min to allow the formation of DNA-liposome complexes. The mixture was added to the cells in 3 ml of serum-free culture medium and incubated at 37°C. After 12 h, the medium was replaced with fresh medium, and 24 h later the cells were harvested. Total RNA was extracted using RNAwiz reagent (Ambion, Austin, TX) and retrotranscribed with poly(dT) primer. To amplify only the messenger derived from the transfections, PCRs were carried out with Ex1-1221 S (5Ј-accaaggacagagacgctggct-3Ј) and Not/Cla rev, (5Ј-tctggacactgcggccgcatcg-3Ј), which is specific for the construct. The conditions used for the PCRs were the following: 94°C for 5 min for the initial denaturation, 94°C for 30 min, 60°C for 1 min, 72°C for 1.5 min for 35 cycles, and 72°C for 7 min for the final extension. The PCRs were optimized to be in the exponential phase of amplification. In some experiments, to quantify the proportion of alternative splicing, cold PCRs were performed for 30 cycles, and then 0.1 l of ␣-32 P-dCTP (1 Ci) was added to each sample along with 1 unit of Taq. The samples were then amplified for 5 more PCR cycles. Radioactive PCR products were run on 8% denaturant acrylamide gel. Phosphorimaging (Instant Imager, Packard Instrument Co.) was used to quantitate PCR amplifications normalized for the cytidine content in the PCR product sequence.
For the SR protein overexpression experiments, 2 g of construct pApo-wt or pApo-A97T were cotransfected with 1 and 2 g of ASF/SF2 and SC35 coding sequences cloned into pCG vector. Each transfection experiment was repeated at least three times.
To generate the different versions of human apoA-II exon 3 ESE RNAs, the following oligonucleotides containing the restriction enzymes sites KpnI and HindIII at the 5Ј and 3Ј splice site, respectively, were used. The sequences were the following: ESEwt, 5Ј-gg tacctatggcaaggacctgatggagaaggtcaagagaagctt-3Ј; ESEA97T, 5Ј-ggtacctatggcaaggacc tgatggtgaaggtcaagagaagctt-3Ј; and ⌬ESE-9, 5Ј-gguaccuauggcaaggaccaggucaagagaagcuu-3Ј. The oligonucleotides were subcloned in the same restriction sites of pBS SK plasmid, linearized with BamHI, and transcribed in vitro in the presence of [␣-32 P]UTP. The UV cross-linking assay was performed by adding the [␣-32 P]UTP-labeled RNA probes (1 ϫ 10 6 cpm/incubation) for 15 min at 30°C with 20 mg of HeLa nuclear extracts (4C Biotech) in 30 ml of final volume. Final binding conditions were 20 mM Hepes, pH 7.9, 72 mM KCl, 1.5 mM MgCl 2, 0.78 mM magnesium acetate, 0.52 mM dithiothreitol, 3.8% glycerol, 0.75 mM ATP, 1 mM GTP, and heparin at a 5g/l final concentration as a nonspecific competitor. The samples were then transferred into the wells of a polystyrene plate and irradiated with UV light on ice (800,000 kJ for 5 min). Unbound RNA was then digested with 30 mg of RNase A and 6 units of RNase T1 (Sigma) by incubation at 37°C for 30 min in a water bath. The samples were then analyzed by 10% SDS-polyacrylamide gel electrophoresis followed by autoradiography.
Immunoprecipitation of SR Proteins following UV Cross-linking-The RNA probes used in this study (ESEwt, ESEA97T, and ⌬ESE-9) were obtained by cloning SalI-EcoRI in Bluescript SKϩ, in which the human apoA-II sequence ranged from intron 2 (nucleotide 1549) to intron 3 (nucleotide 2022), with the wild type ESE mutation A97T ESE and ⌬ESE-9. Each plasmid was then EcoRI-linearized and transcribed in vitro in the presence of [␣-32 P]UTP using T7 RNA polymerase according to standard conditions. The UV cross-linking of labeled RNAs with commercial HeLa nuclear extract was performed as described above. After the 30-min incubation with RNase at 37°C, 150 l of immunoprecipitation buffer (20 mM Tris, pH 8.0, 300 mM NaCl, 1 mM EDTA, 0.25% Nonidet P-40) were added to each sample together with 1 g of monoclonal antibodies and incubated for 2 h at 4°C on a rotator wheel. Anti-SF2/ASF (monoclonal antibody 96) and Anti-SC35 monoclonal antibodies were purchased from Zymed Laboratories (Zymed Laboratories Inc., San Francisco, CA) and Sigma, respectively. Afterward, 30 l of protein A7G-Plus-agarose (Santa Cruz Biotechnology, Santa Cruz, CA) were added to each sample and incubated overnight at 4°C. The beads were subjected to four washing cycles with 1.5 ml of immunoprecipitation buffer and loaded on a 10% SDS-polyacrylamide gel. The gels were run at a constant rate of 30 mA for ϳ3.5 h, dried, and exposed for 4 -6 days with a Biomax Screen (Kodak, Rochester, NY).

RESULTS
Analysis of the ApoA-II Exon 3-As a first approach, we studied the association between the splice site strength and the different splicing behaviors of human apoA-II exon 3 and human CFTR exon 9. This was done by computer in a comparative sequence analysis of the 5Ј and 3Ј splice sites of the human apoA-II exon 3 and human CFTR exon 9.
The strength of these splice sites was calculated according to the Splice Site Prediction by Neural Network program (SSPNN, www.fruit fly.org/seq_tools/splice.html). In general, sequences that have a high score are considered to be strong (the maximum is 1.0 and corresponds to the consensus sequence). The program found the authentic 5Ј and 3Ј splice sites of human CFTR exon 9 (T7 version of the polypyrimidine tract) in which the scores were 0.83 and 0.91, respectively (Fig. 1).
The calculated scores for the human apoA-II exon 3 demonstrated that the authentic donor splice site is ranked as the best possible 5Ј splice site (score ϭ 1.0) within the 600 bp encompassing the genomic region between apoA-II exon 2 and exon 3. On the other hand, the authentic 3Ј splice site of apoA-II exon 3 is not ranked among the possible 3Ј splice sites within the same genomic region (at least with both donor and acceptor cut-off at 0.1). Five cryptic 3Ј splice sites (in which scores ranged from 0.11 up to 0.99) were also identified, (not shown). Overall, these observations suggest that the authentic acceptor splice site of human apoA-II exon 3 might be weak because of the presence of the noncanonical (GU) dinucleotide repeats within the 3Ј splice site.
The apparent weakness of the human apoA-II exon 3 accep-tor splice site (strength Ͻ0.1) in comparison with the human CFTR exon 9 acceptor splice site (strength ϭ 0.91) is in striking contrast with the splicing behavior of the two exons. In fact, human apoA-II exon 3 is constitutively included in mRNA; whereas human CFTR exon 9 is alternatively spliced. Hence, it is plausible that novel accessory cis-acting elements might be involved in the apoA-II exon 3 definition to support its constitutive splicing.
In Vivo System for the Study of ApoA-II Exon 3 Splicing-To study the relevance of the splice site strength and map the possible cis-acting elements involved in the apoA-II exon 3 definition, we generated a eukaryotic gene expression system by cloning the whole 3.2-kbp apoA-II gene including the promoter region and its polyadenylation site (pApo-wt) ( Fig. 2A).
In comparison with the minigene system widely used to study other splicing models (13)(14)(15)(16)(17), where only one exon and its flanking regions are cloned in a heterologous gene context, the apoA-II expression system that we generated contains most of the elements necessary for its transcription and RNA processing, which are also present in the endogenous apoA-II gene. Such a construct should allow the study of the cis-acting elements affecting apoA-II exon 3 splicing in a context as close as possible to the chromosomal background.
The transient transfection of the pApo-wt gene system in Hep3B cells was followed by reverse transcription (RT) using a primer specific for the pApo-wt construct or the endogenous gene. The splicing pattern was then determined by PCR amplification using primers that recognized slightly modified se- quences in the plasmid pApo-wt, which allowed us to discriminate its transcript from the endogenous one. Fig. 2B shows that the splicing efficiency of the apoA-II construct in Hep3B cells was similar to that observed for the endogenous apoA-II gene, displaying about 90% of exon 3 inclusion. An identical experimental strategy was used in all of the following experiments using mutated versions of the pApo-wt gene construct.
In Vivo Effects of (GT) 16 Tract Deletion and Replacement-We have shown previously that when the U tract is removed from human CFTR intron 8, total exon 9 exclusion occurs (10). Moreover, an increase of the number of (GU) dinucleotide repeats reduces the efficiency of exon 9 inclusion. This splicing pattern is in contrast with what is observed in the apoA-II context, where exon 3 inclusion reaches 90%. Therefore, in the context of apoA-II, the (GU) dinucleotide repeats seem to work as a functional polypyrimidine tract. To test this hypothesis, we deleted all the (GT) dinucleotide repeats from the apoA-II intron 2 (Fig. 3A, construct ⌬(gu)). As a control, a sequence containing (CA) 16 dinucleotide repeats was used to replace the (GT) 16 tract maintaining the same length (Fig. 3A,  construct (ca)x(gu)). The deletion and replacement of the (GU) tract resulted in almost 100% of exon 3 being excluded (Fig. 3B,  lanes 2 and 3). This result supports previous studies showing that the removal of all the GU repeats in the apoA-II intron 2 caused ϳ90% of exon 3 to be skipped (11).
To establish if the GU tract is functional or detrimental in systems different from the exons under study (apoA-II exon 3 and CFTR exon 9), we have tested the effect of replacing a canonical polypyrimidine tract with GU repeats. Using a minigene system for the mouse fibronectin EDA exon (12), we found that the substitution of the EDA exon polypyrimidine tract with the apoA-II (GT) 16 tract resulted in 95% of the EDA exon being excluded (Fig. 4).
Altogether these results indicate that the (GU) 16 tract functions as a polypyrimidine tract, which depends on its context and location in the nucleotide sequence. Moreover, these data suggest that the 3Ј splice site definition of the apoA-II exon 3 might be supported by strong and specific cis-acting elements that counteract the effect of the noncanonical polypyrimidine tract.
Mapping of Regulatory Elements within ApoA-II Exon 3-To identify regulatory elements placed within the apoA-II exon 3, we generated hybrid constructs between apoA-II exon 3 and CFTR exon 9 in the same manner as the splicing regulatory elements of CFTR exon 9 had been previously characterized (18).
A set of five CFTR/apoA-II hybrids was generated in which the exon 3 was progressively replaced by CFTR exon 9 and its 3Ј splice site replaced with the allelic configuration (GT) 11 (T) 5 . The constructs were named by indicating the number of apoA-II exon 3 nucleotides replaced by CFTR exon 9 (i.e. 18, 39, 60, 87, and 113) (Fig. 5A).
In comparison with the splicing pattern of pApo-wt, the pCF/Apo18 construct showed 100% of exon 3 inclusion, thus supporting the hypothesis that the CFTR (GT) 11 (T) 5 polypyrimidine tract is stronger than that of the apoA-II (GT) 16 (Fig.  5B, lanes 1 and 2). Subsequently, the pCF/Apo39 showed an increase in the cryptic splice site usage (this cryptic splice site is located 56 nucleotides downstream of the 3Ј splice site of apoA-II exon 3, which is observed as a faint band in apoA-II wild type) (Fig. 5B, lane 3). On the other hand, the pCF/Apo60 constructs showed a new cryptic 3Ј splice site introduced with the extension of the CFTR sequence from nucleotides 39 to 60 (Fig. 5B, lane 4). This cryptic splice site has been characterized previously (10). The transfection of the pCF/Apo87 construct showed 100% inclusion of exon 3, whereas the pCF/Apo113 construct showed 80% exclusion of exon 3 (Fig. 5B, lanes 5 and  6). In principle, the strikingly different splicing patterns of pCF/Apo87 and pCF/Apo113 constructs might be due to the removal of an enhancer element placed within the 26 bp of the apoA-II exon or to the addition of a silencer element introduced within the 26 bp of the CFTR exon 9. However, considering that previous mapping studies did not highlight the presence of any exonic CFTR regulatory element in this 26-bp range (18), we focused attention on the enhancer-like sequence possibly pres-ent in the removed sequences of apoA-II exon 3. To identify the nucleotides within this 26-bp region that contributed to efficient exon 3 splicing, we constructed three 21-bp overlapping deletions encompassing the 26-bp sequence (Fig. 6A). The three constructs (⌬1843-63, ⌬1854 -74, ⌬1866 -86) were transiently expressed in Hep3B, and the RNA was analyzed by RT-PCR.
The ⌬1843-63 deletion caused 80% of exon 3 exclusion, whereas both the ⌬1854 -74 and the ⌬1866 -86 deletions resulted in 55% of exon 3 exclusion (Fig. 6B). Therefore, the region spanning from nucleotide 1843 to 1886 seems to affect positively apoA-II exon 3 definition. In previous studies, a broad spectrum of enhancer-like sequences capable of promoting exon inclusion have been identified (19). Interestingly, the motif GGAGA (Class I of purine-rich enhancers) is also present in the 1856 -1860 nucleotide region of apoA-II exon 3. Moreover, the 1852-1856 region is compatible with the ASF/SF2 motifs found through functional systematic evolution of ligands by exponential enrichment (SELEX) (20). Therefore, we focused attention on the sequence ranging from nucleotide 1852 to 1860 (corresponding to nucleotides 91-99 of exon 3) to be analyzed as a possible regulatory element (Fig. 7A, boxed). To map finely which base(s) in such a region was directly affecting exon 3 definition, point mutations were introduced within this 9-bp region. Each base, one at a time, was replaced by its complementary one, generating nine different constructs that were used to transfect the Hep3B cell line (Fig. 7B). Although most of the substitutions did not have a significant effect on exon 3 processing in Hep3B cells, G92C, A99T, and more noticeably, A97T caused 20, 60, and 85% of exon 3 skipping, respectively (Fig. 7C, lanes 3, 8, and 10).
Interaction of SR Proteins with apoA-II 9-Nucleotide ESE-To identify the trans-acting factors able to bind the FIG. 5. Effect of the progressive apoA-II exon 3 replacement by CFTR exon 9. A, the schemes of the CFTR IVS8-exon 9 (black)/apoA-II exon 3-IVS3 (white) hybrid constructs used for transient transfections in Hep3B cells are shown. The size (in bp) of these replacements is also indicated. Cryptic splice sites of apoA-II exon 3 (56 nucleotides downstream of the acceptor site) and CFTR exon 9 (previously characterized) are shown with a black and white arrowhead, respectively. IVS, intervening sequence. B, denaturing acrylamide gel of radioactive RT-PCRs of the pre-mRNA splicing pattern of the CFTR/apoA-II hybrid constructs. The horizontal arrow on the RT-PCR results indicates aberrant splicing products originated from cryptic 3Ј splice sites located in the hybrid exon 3. The relative amount of exon 3 skipping (ex3Ϫ) was quantified by phosphorimaging analysis of radioactive PCRs as described under "Experimental Procedures." pApo-wt ϭ 10%; pCF/apo113 ϭ 80%. Standard deviations were Ͻ10%. 9-nucleotide enhancer-like sequence within apoA-II exon 3, both EMSA and UV cross-linking assays were performed. For the EMSA experiment, we used in vitro transcribed RNAs including the 9-nucleotide wild type core region of the apoA-II exon 3 (ESEwt) or this same sequence point-mutated (ESE A97T) or deleted (⌬ESE-9) (Fig. 8A). The RNAs were incubated with HeLa nuclear extract in the presence of the nonspecific competitor heparin (see "Experimental Procedures"). As shown in Fig. 8B, ESEwt RNA showed a broad band of shifted material, which is slightly weaker with the ESE-A97T RNA and almost nonexistent with the ⌬ESE-9 RNA (lanes 4 -6). Nevertheless, it should be noted that the region used as the EMSA probe spans the deletion of ⌬1843-63 and ⌬1854 -74. Therefore, these results suggest that several proteins might interact specifically with the wild type apoA-II exon 3 across the 9-nucleotide sequence and its flanking regions, which explains the minimal differences between the ESEwt and ESEA97T complexes observed in the gel shift assay.
The nature of the protein(s) that bind(s) to the wild type apoA-II exon 3 across the 9-nucleotide sequence was then investigated by UV cross-linking of RNA-protein complexes. 32 P-labeled ESEwt, A97T, and ⌬ESE-9 RNA probes were incubated with HeLa nuclear extract and were cross-linked to proteins by exposure to UV light. The resulting 32 P-labeled protein(s) was (were) separated by SDS-PAGE. The pattern of UV crosslinked proteins obtained with these three different constructs did not display any significant difference (data not shown).
To test the possible differential recruitment of SR proteins by the wild type, mutated, or deleted apoA-II 9-nucleotide ESE that may be obscured in the UV cross-linking assay, these were followed by immunoprecipitation with anti-SR protein monoclonal antibodies. The constructs used for these experiments carry ESEwt, ESE A97T, and ⌬ESE-9 (Fig. 9A).
The immunoprecipitation of the UV cross-linked material with monoclonal antibody 1H4 directed against SRp40, SRp55, and SRp75 did not produce any specific immunoprecipitated material (not shown). Instead, the monoclonal antibody 96 anti-SF2/ASF monoclonal antibody (Fig. 9B, left panel) and anti-SC35 monoclonal antibody (right panel) immunoprecipitated specific proteins (lanes 1) following UV cross-linking with nuclear extract of a labeled ESEwt RNA but only traces with the ESE A97T and ⌬ESE-9 control RNAs (lanes 2 and 3). Thus, these results provide evidence that at least two SR proteins are able to interact with the nucleotides across the ESE sequence and are also consistent with the result obtained in the EMSA study, where a clear difference in the RNA-protein complex formation with a construct containing the ESEwt and ⌬ESE-9 is observed.
Effect of ASF/SF2 and SC35 Overexpression on Splicing of ApoA-II Exon 3-To highlight the putative positive effect of ASF/SF2 and SC35, plasmids containing the open reading frame of these SR proteins were cotransfected along with pApo-wt construct. Thus, the possible positive effects of the overexpression of ASF/SF2 and/or SC35 should cause an increase of exon 3 inclusion, even though the percentage of basal exon 3 skipping is low (about 10%). As a control, the construct carrying the point mutation A97T within the ESE was used (pApo-A97T). Fig. 9C showed that the cotransfection of pApo-wt with ASF/SF2 (lane 2) and SC35 (lane 6) promoted exon 3 inclusion when compared with the cotransfection with the empty vector pCG (lanes 1 and 5). Conversely, the cotransfection of ASF/SF2 and SC35 with the construct carrying the A97T point mutation showed no significant variation in levels of exon 3 inclusion (Fig. 9C, lanes 4 and 8). Altogether these results confirm that the disruption of the ESE by the point mutation A97T prevents a functional binding of ASF/SF2 and SC35 to this regulatory sequence and inhibits the positive effect. Hence, both ASF/SF2 and SC35, in which specific interaction to the ESE had already been demonstrated by UV crosslinking/immunoprecipitation (Fig. 9B), have a functionally positive effect, increasing the apoA-II exon 3 inclusion.

DISCUSSION
Human apoA-II exon 3 and its flanking introns share only partially the canonical features of splicing sites. In fact, the 5Ј splice site of intron 3 is well defined and resembles the consen- sus sequences (21), whereas the 3Ј splice site of intron 2 contains a polymorphic sequence made of alternating pyrimidines and purines, instead of the classical polypyrimidine tract.
The comparison of splice sites strength of human apoA-II exon 3 and human CFTR exon 9 has delineated an apparent contradiction between the splicing behavior of the two exons and the strength of the splice sites. Taking advantage of the compact size of the human apoA-II gene, we were able to create an in vivo expression system that includes the whole 1173-bp enhancer promoter region in which the 1226-bp coding region encompasses 4 exons and 3 introns and the 320-bp 3Ј-UTR (where the polyadenylation signal is located). In this way, it was possible to study the exon 3 splicing within a context very similar to the chromosomal context. In addition, a liver-derived cell line (Hep3B) was used to provide a cellular context more similar to that of original tissue where the apoA-II gene is normally expressed. This represents a clear advantage in comparison with the minigene system and is widely used to study other splicing models where only one or a few exons and their flanking regions are cloned in a heterologous gene context.
The first approach toward the characterization of the cisacting elements involved in the apoA-II exon 3 definition led to confirmation of the relevance of the (GU) tract in our system. Both the deletion or the replacement of the (GU) 16 tract with its complementary (CA) 16 tract resulted in 90% exclusion of exon 3 from apoA-II mRNA. It is well known that there is great flexibility in the specific sequence within the polypyrimidine tract. In fact, it has been shown that the introduction of purines into the polypyrimidine tract is detrimental to splicing only if the length of the tract is shortened and if there is a reduction in the number of consecutive uridine residues (4,22,23).
Our in vivo results are consistent also with previous in vitro studies on tracts containing alternating pyrimidine and purine residues, which found that they are functional (5). In contrast, we demonstrated that the replacement of the authentic polypyrimidine tract of the mouse fibronectin EDA exon with the apoA-II (GU) 16 polypyrimidine tract produced EDA exon skipping; therefore, the GU tract is not functionally equivalent in vivo to a polypyrimidine tract in a context different from the apoA-II intron 2.
A possible explanation for the ability of the (GU) repeats to work as a functional polypyrimidine tract derives from the combinatorial action of the cis-acting elements that drive the selection of 3Ј splice sites. It is clear that the strength of the adjacent branch point also plays an important role in splice site selection (3,24). Moreover, studies on regulated alternative splicing have identified exonic splicing elements that facilitate the process of exon definition. Therefore, functional exon definition depends also on the presence of exon enhancers that can clearly rescue splicing from otherwise weak 3Ј splice sites (25)(26)(27)(28). It is now clear that ESEs are not only components of regulated exons but also of constitutively spliced exons (29,30).
The progressive replacement of apoA-II by CFTR sequences led to the identification of a region between nucleotides 87 to 113 of human apoA-II exon 3 as essential for its inclusion in mRNA. On the basis of previous functional studies, which identified different classes of splicing enhancers (20,30), we have outlined a putative enhancer-like sequence located in the region spanning from nucleotide 91 to 99 of exon 3 that was further characterized through mapping by point mutations. Considering the base composition of this 9-bp region (TGATG-GAGA), this sequence would belong to a purine-rich ESE category.
Our UV cross-linking coupled with immunoprecipitation experiments showed the ability of the wild type, but not of the A97T ESE, sequences to bind ASF/SF2 and SC35. The rele-vance of the these two splicing factors for apoA-II exon 3 splicing is confirmed also by in vivo SR protein overexpression experiments. In conclusion, our results indicate that the noncanonical polypyrimidine tract of the apoA-II intron 2 is balanced by the presence of at least one enhancer-like sequence placed within exon 3.
We have shown that ESE is bound by ASF/SF2 and SC35, which, in turn, may have a pivotal role in recruiting the constitutive splicing factor U2AF65, which does not have available a canonical polypyrimidine tract. These data support the hypothesis that the weak polypyrimidine tract of intron 2 might be unable to recruit directly the constitutive splicing factors even in the presence of a strong 5Ј splice site.
A large body of evidence indicates that U2AF35 is required for constitutive splicing and also works as a mediator of enhancerdependent splicing (31)(32)(33). In vitro protein-RNA interaction studies with pre-mRNAs containing either a constitutive or regulated splicing enhancer have shown that U2AF35 directly mediates interactions between U2AF65 and proteins bound to the enhancers (34). Thus, U2AF35 should recruit U2AF65 acting as a bridge between these proteins and the enhancer complex. In the case of apoA-II exon 3, this recruitment should be mediated by the ESE sequence characterized in this paper and its interactions with ASF/SF2 and SC35.